On the number field sieve: polynomial selection and smooth elements in number fields

Abstract

The number field sieve is the asymptotically fastest known algorithm for factoring large integers that are free of small prime factors. Two aspects of the algorithm are considered in this thesis: polynomial selection and smooth elements in number fields. The contributions to polynomial selection are twofold. First, existing methods of polynomial generation, namely those based on Montgomery's method, are extended and tools developed to aid in their analysis. Second, a new approach to polynomial generation is developed and realised. The development of the approach is driven by results obtained on the divisibility properties of univariate resultants. Examples from the literature point toward the utility of applying decoding algorithms for algebraic error-correcting codes to problems of finding elements in a ring with a smooth representation. In this thesis, the problem of finding algebraic integers in a number field with smooth norm is reformulated as a decoding problem for a family of error-correcting codes called NF-codes. An algorithm for solving the weighted list decoding problem for NF-codes is provided. The algorithm is then used to find algebraic integers with norm containing a large smooth factor. Bounds on the existence of such numbers are derived using algorithmic and combinatorial methods.